Draft job descriptions in your brand voice

Consistent, inclusive JDs in minutes.

Claire DuboisBeginnerSaves ≈3h/weekChatGPT
519 views
Free community read · 5 leftUnlock all →

Every hiring manager wrote JDs differently. This gives us one voice and saves the back and forth.

The workflow

  1. Give ChatGPT the role and must-haves
  2. Paste a JD you like as the voice reference
  3. Generate the draft, check for biased language
  4. Hand to the manager for sign-off

Result: On-brand, inclusive JDs without the rewrite cycle.

The prompt

Prompt
Write a job description for [ROLE]. Match the tone of this example: [PASTE]. Include responsibilities, requirements, and a short inclusive close.

A note on what follows. The workflow, the prompt and the result above come from the member. The notes below (why the prompt works, what to watch for, how to adapt it) were written with AI help.

Why this prompt works

Match the tone of this example is doing most of the work here. Describing your voice in adjectives ("friendly but professional") produces the same generic JD every time, while a pasted sample gives the model real sentence length, vocabulary and formatting to imitate. Responsibilities and requirements are named as separate sections because models otherwise merge them into one list where "owns the roadmap" sits next to "5 years experience", which hiring managers cannot review. A short inclusive close is specified as short for a reason: left open, the model writes three paragraphs of boilerplate about diversity that reads as pasted, and the actual role gets pushed below it. The workflow's separate check for biased language pass matters too, since the model reproduces whatever is in your reference JD, bias included.

What to watch for

  • It invents requirements you never gave it: a degree, a years-of-experience floor, tools your team does not use. Every line in the requirements list should trace back to your input.
  • Copying tone from the example also copies its baggage. If your reference JD says "rockstar" or lists ten years for a mid-level role, expect the same in the draft.
  • Compensation, location and work-authorization language get hallucinated or omitted, and several jurisdictions require a salary range. Fill those in yourself, never from the draft.
  • Requirement inflation quietly narrows the funnel. Ask which requirements are genuinely mandatory before the manager signs off.

How to adapt it

  • Split the bar explicitly: Separate requirements into must-have and nice-to-have. Keep must-have to five items or fewer.
  • For regulated or pay-transparent markets, add Include a salary range of [RANGE] and a location line stating [POLICY]. Do not invent either.
  • To run the bias check in the same pass, add After the draft, list any gendered, ageist or exclusionary wording you used and a neutral replacement.

What good output looks like

A good draft reads like your existing JDs, with the same section order and roughly the same length, and every requirement maps to something you actually told it. Five or fewer must-haves, a close that is two or three sentences, no invented degree. If the draft is noticeably longer than your reference or opens with a paragraph about the company's mission you never supplied, the tone sample is being ignored, so paste a shorter example and re-run.

Want to share your own workflows?

AI Academy members publish workflows, vote, and get featured in Techpresso.

Become a member

Discussion

Loading…